The shape of the TF looks nice but the calibration must be wrong.
Suppose 1/f slope with 10^-4 rad/V at 100kHz. i.e. m_pm = 10/f rad/V
This means m_fm = 10 Hz/V. This is 10^6 times smaller than that of LWE NPRO.
(Edit: Corrected some numbers but it is not significant)
Kiwamu and I measure the PZT response of the Innolight this evening from 24 kHz to 2MHz.
We locked the PLL at ~50 MHz offset using the Lightwave NPRO and and swept the Innolight with the network analyzer (using the script I made; it has one peculiar property, but it does work correctly).
We will post the plot of the Lightwave PZT response tomorrow morning.